I agree with the previous responses. The purpose of the investigation is to define the root cause(s). The purpose of the corrective action is to address that/those identified root cause(s). Unless lack of training is identified as your root cause, retraining should rarely be used as a corrective action. And if it is, that opens up a whole other issue, where a CAPA should be used to address why there was a lack of training in the first place (unless that was also identified in the deviation). I review many deviations/non-conformances from many different organizations (we outsource work), and typically we see training as a band aid response. If there is training in place already, why was that initial training not effective? Why would retraining be any more effective? Would those who are not part of the "retraining" (i.e. new hires) be aware of the issue identified in the the deviation/non-conformance, or will the issue be repeated if only trained using the same initial training method? Tribal knowledge is never a robust method for transferring information, so if the expectation is that you can reinforce something by repeating it that isn't clear initially, from an external perspective, that seems dubious. As a side note, human error should also rarely be a root cause. If people are making mistakes, it probably isn't because they want to. There is probably something with the process that is not clear, or a hindrance of some sort that should be identified and improved to ensure they are able to perform without error. That would be the actual root cause. Again, this is why retraining is frowned upon. It implies the true root cause has not been identified.
When we push back on training as a CA, it is often because we have seen a repeat occurrence of the deviation elsewhere, or see the risk of a repeat when there is not an additional change implemented to the system.